Release a subproject to PyPI

1. Check latest version on PyPI

Before bumping or publishing, confirm the current published version so the new release is higher.

Option A – pip (any subproject):

pip index versions <package-name>

Example: pip index versions lightrag-memgraph → shows e.g. 0.1.4.

Option B – PyPI JSON API:

curl -s https://pypi.org/pypi/<package-name>/json | python -c "import sys,json; d=json.load(sys.stdin); print(d['info']['version'])"

Use the same package-name as in pyproject.toml (e.g. lightrag-memgraph).

2. Bump version if needed

In the subproject’s pyproject.toml, set version to a value greater than the latest on PyPI (e.g. patch: 0.1.4 → 0.1.5).

3. Build and publish

From the repo root (ai-toolkit), source env then run from the subproject directory:

source .env
cd <subproject-path>   # e.g. integrations/lightrag-memgraph
uv build
uv publish

Build output location: uv build may write artifacts to the top-level ai-toolkit dist/ directory rather than the subproject’s. If uv publish reports "No files found to publish", run publish from the ai-toolkit root and pass the built files explicitly:

cd /path/to/ai-toolkit
source .env
uv publish dist/<package>_<normalized>-<version>-*.whl dist/<package>_<normalized>-<version>.tar.gz

Example for lightrag-memgraph 0.1.4: uv publish dist/lightrag_memgraph-0.1.4-py3-none-any.whl dist/lightrag_memgraph-0.1.4.tar.gz

Requires UV_PUBLISH_TOKEN in .env (and UV_PUBLISH_USERNAME for CI; local token auth may not need it).

Subproject paths (from ai-toolkit root)